not applicable
not applicable
not applicable
1. Exceptional Returns: The Economic Value of America""s Investment in Medical Research, report by the Funding First initiative of the Mary Woodard Lasker Cheritable Trust, Washington, D.C., May 2000 (http://www.laskerfoundation.org/fundingfirst)
2. R. Marbach et al., Non-invasive Blood Glucose Assay by Near-Infrared Diffuse Reflection Spectroscopy of the Human Inner Lip, Appl. Spectrosc. 47, 875-881 (1993)
3. R. Marbach, On Wiener Filtering and the Physics Behind Statistical Modeling, to be published in the Journal of Biomedical Optics (accepted Jul. 6, 2001)
4. R. R. Alfano and S. G. Demos, Imaging of Objects Based Upon the Polarization or Depolarization of Light, U.S. Pat. No. 5,847,394 filed Aug. 28, 1996
5. Y. Maekawa et al., Non-invasive Blood Analyzer and Method Using the Same, U.S. Pat. No. 5,769,076 filed May 2, 1996
6. D. Hochman and M. M. Haglund, Optical Imaging Methods, U.S. Pat. No. 5,845,639 filed Nov. 11, 1994
7. R. Marbach and H. M. Heise, Optical Diffuse Reflectance Accessory for Measurements of Skin Tissue by Near-Infrared Spectroscopy, Appl. Optics 34, 610-621 (1995)
The invention relates to methods and apparata for improving the tracking accuracy and signal-to-noise ratio of noninvasive blood analysis methods.
Recent years have seen significant efforts spent on developing methods that can analyze human blood noninvasively as well as with sufficient accuracy, speed, low cost, minimal discomfort to the patient, and at the point-of-care. The biggest market segment for noninvasive blood analyzers is the diabetes market, because the disease affects a significant fraction of the population and patients are required to perform regular and frequent measurements of their blood glucose concentration. The following discussion will therefore concentrate on glucose as the primary candidate to which this invention can be applied, however, this is only meant in an exemplary way since the invention can be applied equally well to noninvasive measurement methods of other blood constituents, e.g., urea.
A conservative estimate by this author is that  greater than US$ 2 billion have been spent during the last decade on RandD expenses for an accurate noninvasive blood glucose monitor. The reason, of course, is the significant market size and the potential ease with which the existing fingerprick devices could be pushed out of the market by even a fairly expensive noninvasive monitor if only the noninvasive device was accurate enough. Government also has a strong interest in an accurate noninvasive monitor because of the expected decrease in diabetes-related health care costs, which are currently estimated at $92 billion annually in the US [1].
Many different methods for the non-invasive measurement of blood glucose and other blood components-have been proposed. Virtually all are based on optical measurement techniques, i.e., they measure changes in the, e.g., absorbance, scatter, fluorescence, emission, polarization, Raman scatter, or a combination of these effects; in a tissue as a function of the glucose concentration in the blood. Further differences come from the different proposed wavelength regions of the electromagnetic spectrum and locations on the body. Wavelength ranges proposed range from the ultraviolet (xcex less than 400 nm) to the far infrared (xcex greater than 20,000 nm) and typical locations proposed include the volar forearm, lip, fingertip, ear lope, and eye. Many of the published claims must be judged with extreme caution, especially in cases when the basic physical relationships are unclear or when the published data is statistically grossly insufficient.
The most promising noninvasive methods are absorbance-based optical measurements performed in diffuse reflection geometry in the near-infrared wavelength region (NIR). Proof of the basic technical feasibility was published in 1993 [2]; however, accuracy was insufficient at about 50 mg/dL root-mean-square (RMS) of measurement error, which is about 3 times larger than the clinically required value. Surprisingly, almost 10 years and $2+billion later, accuracy has not improved substantially since. We will now disclose the reason behind the limitation to accuracy and then, in the descriptive part of this text go on and disclose a method and apparata to overcome it.
The following discussion will concentrate on NIR measurements because these methods have the best chances for commercial success and are therefore prime candidates to which this invention can be applied. However, again, this is not meant in an exclusive way. In fact, the invention can be applied equally well to other noninvasive measurement techniques, based on other optical or even non-optical methods, because the problem solved by the invention applies equally to all noninvasive measurement methods. In the following, whenever words like xe2x80x9coptical spectrum,xe2x80x9d xe2x80x9coptically probed skin volumexe2x80x9d etc. are used, they are meant only in an exemplary way.
The accuracy of all non-invasive methods is affected by two types of error, viz. (a) the xe2x80x9cspectral errorxe2x80x9d due to the noise generated by the hardware of the noninvasive device, its sampling interface, and the interfering spectra from the other blood and tissue components and (b) the xe2x80x9ctracking errorxe2x80x9d generated by the fact that the glucose concentration in the probed skin volume is not perfectly correlated with the glucose concentration in the blood. The latter type of error occurs because the glucose concentration in the probed skin volume (PSV) is an average of the glucose dissolved in the interstitial fluid (ISF) and the glucose in the blood. The instantaneous glucose concentration in the ISF (ISFG) can be very different from the glucose concentration in the blood (BG) because of the complicated temporal and spatial relationships between glucose intake and transport, and insulin intake and transport, in the body of a diabetic.
The accuracy of all non-invasive methods is judged by comparison to a high-quality invasive method, which serves as a secondary standard and calibration reference to the noninvasive method. Thus, even if one assumed that both the spectral error of the noninvasive device was zero (i.e., it measured glucose in the PSV (PSVG) with 100% accuracy) and the error of the invasive standard device was zero (i.e., it measured BG with 100% accuracy) then there would still be the difference between the PSVG and the BG causing a difference between the two devices. This is the tracking error, which is counted as an xe2x80x9cerrorxe2x80x9d of the noninvasive device, because the value of the invasive reference method is assumed to be xe2x80x9ctruexe2x80x9d by definition. A detailed description of how the tracking error and the spectral error interact and combine to affect the overall measurement accuracy has recently become available [3].
Describing the situation in terms of time functions, it can be said that ISFG is virtually always lagging behind BG when BG goes up. When BG goes down, however, the ISFG in diabetic patients can either be lagging or leading, depending on the status of the complicated push-pull mechanism that controls the ISFG in the PSV. The exact time relationship between BG and ISFG is unpredictable in diabetics and can not be described with just a single number for xe2x80x9clag time.xe2x80x9d If one were to plot typical daily time profiles of diabetic BG and ISFG into a single graph and ask people to visually estimate the average time offset, numbers as high as 1 hour would occur commonly, and 2 hours occasionally. Medical doctors are primarily interested in BG and not in ISFG because today""s invasive methods measure BG. The bottom line is that in diabetics, ISFG does not track BG closely enough to allow any of today""s noninvasive methods to achieve full clinical usefulness and to successfully pass comparisons with invasive methods.
The fact that many of today""s NIR absorbance-based optical methods are limited by tracking error and no longer by spectral error, can be rigorously proven by using the theory published in [3] and is also evidenced by the fact that measurement precision (repeatability) is often much better than the overall long-term accuracy. In a nutshell, modern NIR methods have become good at measuring the wrong thing. In order to improve performance, it is therefore necessary to improve upon the measurement method itself, because further improvements to the hardware alone will have virtually no effect on the measurement accuracy a.k.a. clinical usefulness.
It is obvious that an invention that solves the glucose accuracy problem can also be applied to improve the accuracy of noninvasive measurements of other blood components. E.g., during treatment of a dialysis patient, one can measure his PSV-urea concentration noninvasively and in real-time, e.g., by NIR diffuse reflection spectroscopy. The invention disclosed below can be applied to improve the accuracy of the urea measurement for the exact same reason that was discussed above, viz., to improve the tracking accuracy between the urea concentration in the blood and the urea concentration in the optically probed skin volume.
Three further remarks are on order. First, the invention disclosed below can be used in conjunction with all noninvasive measurement methods, including existing ones. Important changes have to be made to existing pieces of noninvasive device hardware, however, to (1) optimize the existing hardware so it can realize the full potential of the accuracy advantage provided by this invention, see the detailed discussion below, and (2) to accommodate the additional apparatus necessary for this invention. Second, in order for this invention to provide an accuracy advantage, the noninvasive method that it is applied to must itself be limited by tracking error and no longer by spectral error. In other words, if your precision is not yet better than your accuracy, then adding this invention will not help you. Thirdly and most importantly, because the correlation coefficient between the PSVG and BG is low at typically 0.85, the accuracy is just starting to become a steep function of correlation coefficient. In other words, every little bit of improvement in correlation coefficient is just starting to really pay off.
This invention provides methods and apparata for improving the tracking accuracy and signal-to-noise ratio of noninvasive blood analysis methods. The correlation between the component concentration in the probed skin volume and the component concentration in the blood is improved by selecting particular locations on the patient""s skin which provide a significantly higher density of capillary vessels than found on average (xe2x80x9csweet spotsxe2x80x9d). The higher capillary density causes the component concentration in the probed skin volume to better track the component concentration in the blood and, as a welcome side effect, also improves the signal-to-noise ratio (SNR) of the noninvasive measurement method itself.
Methods for locating sweet spots and selecting them for measurement are described. These methods are based on low-cost, real-time optical imaging apparatus using visible wavelengths, and work best when applied to particular locations on the body, viz., mucousas. Also described are several embodiments of sweet spot imaging noninvasive measurement systems that integrate the low-cost optical imaging of capillaries in the visible wavelength range, with the high-accuracy noninvasive measurement in the component-specific wavelength range, e.g., the near-infrared.
An important characteristic of the sweet spot method is its low added cost. This is due to two fortunate facts. First, there are ways to use mass-produced CCD or CMOS cameras to generate real-time capillary images of sufficient quality. And second, it is possible to locate sweet spots on the skin that are large enough to overcome the spatial resolution limit set by the optical scattering in the skin. In other words, because sweet spots with lateral sizes in the range from about 0.5 to 3 mm can be found, standard techniques as applied in existing noninvasive measurement methods, e.g., NIR diffuse reflection, can be modified to achieve sufficient spatial resolution to allow selective probing of sweet spots. Expensive methods for increasing the spatial resolution of the noninvasive measurement like, e.g., optical coherence tomography or time-of-flight gating, can be avoided.